Over these past two months, we have discussed extensively the impact generative AI has on how we write, speak, learn, and teach. Behind this series, was several more months of research to understand the overall themes of existing literature in this field of education and pedagogy, and perhaps several more to commit myself to learning a language using generative artificial intelligence (GenAI). As much fun as it was to do my research on these topics, I must admit that I could not bring myself to use nor integrate GenAI in how I learn languages. And so, today, to cap off this series, I want to share what I have learned, and my experiences behind the scenes.

Starting off, I particularly enjoyed reading up and getting up to speed with literature surrounding the state of GenAI in language learning. Prior to this, I was aware of the adverse impacts of reliance on GenAI on the development of critical thinking and other higher cognitive skills, but uncertain on its effects on more basic or foundational skills in foreign language acquisition. It is definitely heartening to know that there is an increasing academic interest in this field of research, and I am optimistic that we will build a better understanding over time, and even uncover some of the long-term effects as well.

Perhaps the most interesting articles I covered in this series were about the impacts of GenAI on the words we use. GenAI is notorious for its verbose but substance-lacking way of speech, as its responses are usually stuffed with style words. Before reading deeper into this topic, I had the suspicion that constant contact with GenAI would affect word preferences in humans, but large-scale empirical studies would be needed to truly elucidate these effects (or lack thereof). These articles have, to a large part, agreed with my suspicions, though I would like to see further replicative studies or other kinds of empirical studies (such as in non-English languages) to see if this phenomenon is really happening, or if this phenomenon is unique to English.

These articles also made me do a retrospection on the way I write, and how my writing style has changed over time. Before learning about the use of GenAI, I used to mention words like ‘dive into’, ‘deep dive’, ‘delve’, and ‘explore’ when writing my introductions and concluding paragraphs on here. These days are definitely long gone, and I have developed an aversion to using these words in my writing. In the post-GenAI release era, which started around late-2022, this aversion simply grew, and eventually, I just had to come up with alternatives to distinguish my writing from those of GenAI responses. To say that GenAI did not have an impact on my word preferences would be wildly inaccurate, as I am motivated to write like myself, and not some bot.

I think the part that carried the largest inertia was the attempt at integrating GenAI into my language learning experience for review purposes. Part of this may come from a discrepancy in the language that I am learning this year, compared with what GenAI is capable of with regards to that language. As such, I had a greater reluctance in choosing and using the application to review, which in retrospect, I could perhaps better thought of a plan if my target language was part of the relatively more popular foreign languages. This is, of course, in spite of the systematic review I had gone through earlier, which suggested that there is some empirical evidence suggesting that integration of chatbots into the mix of language learning methods could be beneficial for language learning. In my defense, this is a novel field of research that has the potential to pose as a paradigm shift in how languages are learned, and one cannot rule out the possibility of novelty bias in studies like these. I would hold out on these sorts of GenAI-integrated methods until more research is done to build a more established evidence base.

The other difficulty I encountered was selecting a reputable language learning application that integrates or centers around the use of GenAI. There were a couple of routes I could take, such as those that are featured using influencer marketing, and those that appear to top the language learning applications charts in the App Store. I did eventually settle on a couple of applications to review, which I really tried maintaining an open mind to do an unbiased review. Admittedly, it was a trial of how well I could maintain this attitude in spite of the overall negative reputation GenAI has, and my own personal biases against GenAI. Looking back, I think it was worth it to step wildly off my comfort zone in my language learning methods (like language immersion), though these are applications I would not be touching again in future. Perhaps these applications could work for other learners, but definitely not me for now and the foreseeable future. I guess this is why I mentioned ‘biting the bullet’ back in June, before starting this series of posts related to GenAI in linguistics and language learning.

In any case, I wound up not enjoying my experience on Airlearn at all, as upon first sight, I was quite appalled by the slop art that featured heavily throughout the application, and an AI-generated voiceover that was wildly inaccurate at times. In my review, I called Airlearn a warning to Duolingo against fully pivoting to GenAI in language course design, and I stand by this stance. Airlearn really worked like a cheap and shoddy version of what I remember Duolingo to be, and if one does not really learn much from Duolingo, they would learn even less from Airlearn. I did try to pick out some positives, such as the diversity of Indian languages covered by Airlearn, but these are definitely overwhelmingly outweighed by the quality issues in what is a slop application. This resulted in my strongly negative opinion on Airlearn, and despite them having sponsored some channels before, I would not recommend picking up Airlearn at all.

Given my bad impression of Airlearn, I wanted to provide better alternatives to these ‘innovative’ language learning methods, and picked the topic of overcoming language or speaking anxiety to work on. In my review of language cafés, I wanted to provide a positive spin to the overall grim outlook on AI use in language learning, that there is something substantially and drastically better than just talking to a chatbot powered by a large language model. Emphasising on the human connection, and the general interest to improve mastery of the participants’ respective target languages, I hoped to show that these in-person or online interaction sessions would develop interpersonal skills and other soft skills more so than huddling up in an AI-powered chatbot parasocial bubble. In this day and age, I definitely see the importance of genuine connection, and I hope that my arguments have swayed readers away from considering using a chatbot, and towards looking for language learning communities online or in their vicinity. I think that the review on language cafés would have been more meaningful than just reviewing a chatbot for which there are many similarly-functioning ones, and I genuinely enjoyed writing up this review way more than that for Airlearn.

This has been my final thoughts about the series that I have just done. I have pretty much gone through my thoughts about what I enjoyed the most, the least, and found the most interesting when researching and writing this series over the past months, and I would like to thank you for following this little arc of essays in this very topic. I might want to revisit this field of linguistics at some point in the future, since the GenAI ecosystem is still pretty dynamic despite the increasing notoriety and overall negative reputation these programmes are receiving. In the meantime, I will remain on the hunt for more studies in the impacts of GenAI in the languages we speak and how we learn languages.